Virtual servers and storage

Are the virtual server and disaster recovery/business continuance management roles mutually exclusive?

 Andi Mann: This is a big area of overlap for server administrators, so they are not mutually exclusive, but the two groups have to work together. Many organizations – in our research, around 60% – are looking at virtualization as a way to improve their DR/BC. So IT managers are asking the server or virtualization team to deploy high availability, live migration, dynamic resource scheduling and other virtualization technologies to build DR plans around. However, these server admins, especially at larger organizations, don't have the skills, knowledge, or even the time to become experts in the storage side of data recovery and high availability.


 Curtis Breville: At the SMB level, the person responsible for the front-end servers and back-end processes such as backup, recovery, replication and many of the DR/BC may be the same person, or a small group of people who work closely together. In the largest enterprises, these roles are more defined and separated. Though it would be ideal that both the server and the storage sides know much more about one another, the fact that each normally has dozens of projects they are working on and very little downtime to pick up a new set of skills and disciplines will keep them separated for some time.

When does the backup/storage administration team need to be in contact with the server and virtual server team?

AM: From the server administrator perspective, this needs to be an ongoing dialog. Starting with the planning phase, the server team needs to know that the storage team is going to handle the new throughput requirements of virtualization. Then during initial implementation, making sure the VMs and hosts are connected to the right LUNs is critical. If the implementation includes dynamic provisioning, VMs need to be able to maintain continuous connections with the image storage. If the site is using live migration, the VMs need continuous connections to the application components and data, even when workloads are moving from one server to another (and even from one data center to another). There has to be good communication about backups too, because virtualization introduces new challenges and requirements, like the need to back up the VM host, or making sure backups running in one VM don't slow down another VM that is sharing the same physical box. It requires an ongoing dialog.

CB: Large enterprises generally have rules and guidelines in place regarding changes taking place in production environments. Requests need to be made, forms need to be filled out, and the resources, such as disk space and performance, are provided in advance. This change management process, in general, assures everybody has what they need. Weekly or bi-monthly meetings with people from the server side, applications, database, and storage groups are scheduled to discuss challenges in the ability to deliver what is needed. In these large environments, not much more needs to be done. However, in the smaller environments that are less formal and have people with multiple roles and changes being made more often, there are daily discussions going on between the backup/storage administrators and the server and virtual server people.

Do storage administrators need to be more involved with the server side, or vice versa?

CB: Storage administrators will provide the storage to meet the needs of the server side and the applications. Most storage administrators usually do not care if the server is physical or virtual. Most servers today are connected to or communicated with by their logical connection to the network, so whether the server is virtual or physical is not so important from the storage side. The server/virtual administrators are usually the ones requesting resources and it is better for that side to know how much physical disk is available and what sort of protection, recovery, and back-end processes are in place to support what they want to do.

AM: The skills burden in virtualization is more on the server side, so server administrators probably need to be more involved in storage than the other way around. For most operations, a virtual server looks pretty much like a physical server, from the outside – so storage operations are mostly transparent. However, when you get to the functions that are fundamentally internal to the virtualization process - such as provisioning, live migration, or VM backup and recovery - that is where the server team is going to need the expertise of the storage team.

Should the restores of the server level be performed by the server team or the storage/backup team?

AM: This again is really a joint effort, but it depends on the technologies that are in use. If it is just a full system restore, from a backup of the virtual image files, then it's straightforward. However, if you start to get into low-level restores – at the file system or block level – then this is where the storage team needs to get involved. By the same token, when the storage team needs to use specialized tools for virtualization backup and restore – like VMware's VCB or third-party tools – they need to make sure the server team is involved to make sure the restores go on the right VMs and the right hosts.

CB: There are going to be times in which actions can be taken from the user-interface for restores. More complex restorations may require data to come back from tapes. Where the problem generated from in the first place, that caused the need for a restore, will also determine who needs to be involved.

Server virtualization vendors are making it easier for server administrators to replicate data, do bare metal restores, and build new virtual servers from previous images stored. Is this a good thing or bad thing for the storage/backup team?

CB: It is a very interesting time with this VM sprawl as new data protection solutions focused on the virtual side are being introduced almost weekly. Lefthand Networks' Virtual SAN Appliance, BakBone's NetVault, and Continuity Software's RecoverGuard are just a few of the many new data protection solutions targeted at the virtual server space. Consideration of these can come from either the server or storage side, but should involve both to make sure they are right for the environment.

AM: This is both good news and bad news for storage and backup teams. Virtualization makes storage management so much more critical, so it is a lot more work to make sure the storage connections are right, the backups are workable, the SAN or NAS is accessible to all the right VMs, etc. This requires new skills too, which take time and effort to build up. But as storage administration becomes even more critical to the virtual infrastructure, and to successful business service delivery, so the administrators themselves become more important too. And with VM sprawl affecting most enterprises, the time and effort saved on routine tasks such as storage provisioning and system restores will be much better spent on strategic storage management activities such as implementing de-duplication and other efficiency measures to cope with the server sprawl.

Curtis Breville is the senior analyst of storage management, and Andi Mann is a vice president and leader of the systems and storage management practices, at Enterprise Management Associates (EMA), www.enterprisemanagement.com. Curtis can be contacted at cbreville@enterprisemanagement.com and Andi can be contacted at amann@enterprisemanagement.com.

This article was originally published on May 01, 2009